Shoulder MRI accuracy in the community setting

Abstract

Hypothesis/background: Evaluate the accuracy for shoulder magnetic resonance imaging (MRI) interpretations of typical community radiologists and compare this to selected fellowship trained musculoskeletal radiologists.

Methods: The MRIs of 104 shoulders of 100 consecutive different patients who had undergone shoulder arthroscopy by a single surgeon were included in this prospective study. The 104 typed MRI reports of community radiologists were examined, and the positive or negative findings were recorded for the anterior, superior, and posterior labra, and the long head biceps tendon and rotator cuff. The 104 MRIs were then interpreted by 2 fellowship trained musculoskeletal radiologists who were blinded to the nature of this study. They were asked to specifically comment on the same shoulder pathologies and to subjectively rate the quality of each MRI study using a visual analogue score (VAS).

Results: The shoulder pathologies documented by arthroscopy included rotator cuff tears (55 full, 10 partial), labral tears (12 anterior, 17 posterior, 29 superior), and 31 biceps abnormalities. There were 69 noncontrast MRIs and 35 MRI arthrograms. Comparing the community radiologists' accuracies with the 2 selected musculoskeletal radiologists' accuracies, there were no significant differences for any of the shoulder pathologies studied. There was a significant positive correlation between VAS and accuracies for both musculoskeletal radiologists in diagnosing posterior labral tears only. The use of contrast in these MRIs only significantly improved both of the musculoskeletal radiologists' accuracies in diagnosing biceps lesions.

Conclusion: The accuracies for shoulder MRI in this community setting were not improved by having the MRIs interpreted by selected fellowship trained musculoskeletal radiologists.